Python Pandas非相等联接

Question

有表

import pandas as pd
import numpy as np

list_1=[['Steven',np.nan,'C1'],
        ['Michael',np.nan,'C2'],
        ['Robert',np.nan,'C3'],
        ['Buchanan',np.nan,'C1'],
        ['Suyama',np.nan,'C2'],
        ['King',np.nan,'C3']]
labels=['first_name','last_name','class']
df=pd.DataFrame(list_1,columns=labels)
df

OUT

    first_name  last_name   class
0   Steven       NaN         C1
1   Michael      NaN         C2
2   Robert       NaN         C3
3   Buchanan     NaN         C1
4   Suyama       NaN         C2
5   King         NaN         C3

需要：

first_name  last_name
Steven       Buchanan
Michael      Suyama
Robert       King

所以我需要进行非平等的加入等效的SQL查询：

;with cte as
(
SELECT first_name,
        class,
        ROW_NUMBER() OVER (partition by class ORDER BY first_name) as rn
FROM students
)
select c_fn.first_name,
        c_ln.first_name
from cte c_fn join cte c_ln on c_fn.class=c_ln.class and c_ln.rn< c_fn.rn

或作为SQL查询：

;with cte as
(
SELECT first_name,
        last_name,
        ROW_NUMBER() OVER ( ORDER BY (select null)) as rn
FROM students
)
select fn.first_name,
        ln.first_name as last_name
from cte fn join cte ln on ln.rn=fn.rn+3

PANDAS中的问题是，不能用MERGE完成非均等的自联接。而且我找不到其他方法.....

Answer 1

我们可以通过将groupby与agg结合使用并以字符串形式连接，从而以更智能的方式解决熊猫问题。然后我们split他们到列：

dfn = df.groupby('class').agg(' '.join)['first_name'].str.split(' ', expand=True)
dfn.columns = [df.columns[:2]]
dfn = dfn.reset_index(drop=True)

  first_name last_name
0     Steven  Buchanan
1    Michael    Suyama
2     Robert      King

Answer 2

您可以将索引设置为'class'并选择各个名称：

df = df.setIndex('class')
first_name = df.loc["C1", "first_name"].values[0]
last_name = df.loc["C1", "last_name"].values[1]